2 * Copyright (C) 1995 Linus Torvalds
4 * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
6 * Memory region support
7 * David Parsons <orc@pell.chi.il.us>, July-August 1999
9 * Added E820 sanitization routine (removes overlapping memory regions);
10 * Brian Moyle <bmoyle@mvista.com>, February 2001
12 * Moved CPU detection code to cpu/${cpu}.c
13 * Patrick Mochel <mochel@osdl.org>, March 2002
15 * Provisions for empty E820 memory regions (reported by certain BIOSes).
16 * Alex Achenbach <xela@slit.de>, December 2002.
21 * This file handles the architecture-dependent parts of initialization
24 #include <linux/sched.h>
26 #include <linux/mmzone.h>
27 #include <linux/screen_info.h>
28 #include <linux/ioport.h>
29 #include <linux/acpi.h>
30 #include <linux/apm_bios.h>
31 #include <linux/initrd.h>
32 #include <linux/bootmem.h>
33 #include <linux/seq_file.h>
34 #include <linux/console.h>
35 #include <linux/mca.h>
36 #include <linux/root_dev.h>
37 #include <linux/highmem.h>
38 #include <linux/module.h>
39 #include <linux/efi.h>
40 #include <linux/init.h>
41 #include <linux/edd.h>
42 #include <linux/iscsi_ibft.h>
43 #include <linux/nodemask.h>
44 #include <linux/kexec.h>
45 #include <linux/dmi.h>
46 #include <linux/pfn.h>
47 #include <linux/pci.h>
48 #include <linux/init_ohci1394_dma.h>
49 #include <linux/kvm_para.h>
51 #include <video/edid.h>
56 #include <asm/mpspec.h>
57 #include <asm/mmzone.h>
58 #include <asm/setup.h>
59 #include <asm/arch_hooks.h>
60 #include <asm/sections.h>
62 #include <asm/io_apic.h>
66 #include <setup_arch.h>
67 #include <asm/bios_ebda.h>
68 #include <asm/cacheflush.h>
69 #include <asm/processor.h>
73 /* This value is set up by the early boot code to point to the value
74 immediately after the boot time page tables. It contains a *physical*
75 address, and must not be in the .bss segment! */
76 unsigned long init_pg_tables_start __initdata = ~0UL;
77 unsigned long init_pg_tables_end __initdata = ~0UL;
82 static struct resource data_resource = {
83 .name = "Kernel data",
86 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
89 static struct resource code_resource = {
90 .name = "Kernel code",
93 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
96 static struct resource bss_resource = {
100 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
103 static struct resource video_ram_resource = {
104 .name = "Video RAM area",
107 .flags = IORESOURCE_BUSY | IORESOURCE_MEM
110 /* cpu data as detected by the assembly code in head.S */
111 struct cpuinfo_x86 new_cpu_data __cpuinitdata = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
112 /* common cpu data for all cpus */
113 struct cpuinfo_x86 boot_cpu_data __read_mostly = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
114 EXPORT_SYMBOL(boot_cpu_data);
116 unsigned int def_to_bigsmp;
118 #ifndef CONFIG_X86_PAE
119 unsigned long mmu_cr4_features;
121 unsigned long mmu_cr4_features = X86_CR4_PAE;
124 /* for MCA, but anyone else can use it if they want */
125 unsigned int machine_id;
126 unsigned int machine_submodel_id;
127 unsigned int BIOS_revision;
129 /* Boot loader ID as an integer, for the benefit of proc_dointvec */
136 char dmi_alloc_data[DMI_MAX_DATA];
141 struct screen_info screen_info;
142 EXPORT_SYMBOL(screen_info);
143 struct apm_info apm_info;
144 EXPORT_SYMBOL(apm_info);
145 struct edid_info edid_info;
146 EXPORT_SYMBOL_GPL(edid_info);
147 struct ist_info ist_info;
148 #if defined(CONFIG_X86_SPEEDSTEP_SMI) || \
149 defined(CONFIG_X86_SPEEDSTEP_SMI_MODULE)
150 EXPORT_SYMBOL(ist_info);
153 extern int root_mountflags;
155 unsigned long saved_video_mode;
157 #define RAMDISK_IMAGE_START_MASK 0x07FF
158 #define RAMDISK_PROMPT_FLAG 0x8000
159 #define RAMDISK_LOAD_FLAG 0x4000
161 static char __initdata command_line[COMMAND_LINE_SIZE];
163 #if defined(CONFIG_EDD) || defined(CONFIG_EDD_MODULE)
165 #ifdef CONFIG_EDD_MODULE
169 * copy_edd() - Copy the BIOS EDD information
170 * from boot_params into a safe place.
173 static inline void copy_edd(void)
175 memcpy(edd.mbr_signature, boot_params.edd_mbr_sig_buffer,
176 sizeof(edd.mbr_signature));
177 memcpy(edd.edd_info, boot_params.eddbuf, sizeof(edd.edd_info));
178 edd.mbr_signature_nr = boot_params.edd_mbr_sig_buf_entries;
179 edd.edd_info_nr = boot_params.eddbuf_entries;
182 static inline void copy_edd(void)
187 #ifdef CONFIG_BLK_DEV_INITRD
189 static bool do_relocate_initrd = false;
191 void __init reserve_initrd(void)
193 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
194 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
195 u64 ramdisk_end = ramdisk_image + ramdisk_size;
196 u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
199 if (!boot_params.hdr.type_of_loader ||
200 !ramdisk_image || !ramdisk_size)
201 return; /* No initrd provided by bootloader */
205 if (ramdisk_size >= end_of_lowmem/2) {
206 free_early(ramdisk_image, ramdisk_end);
207 printk(KERN_ERR "initrd too large to handle, "
208 "disabling initrd\n");
212 printk(KERN_INFO "old RAMDISK: %08llx - %08llx\n", ramdisk_image,
216 if (ramdisk_end <= end_of_lowmem) {
217 /* All in lowmem, easy case */
219 * don't need to reserve again, already reserved early
220 * in i386_start_kernel
222 initrd_start = ramdisk_image + PAGE_OFFSET;
223 initrd_end = initrd_start + ramdisk_size;
227 /* We need to move the initrd down into lowmem */
228 ramdisk_here = find_e820_area(min_low_pfn<<PAGE_SHIFT,
229 end_of_lowmem, ramdisk_size,
232 if (ramdisk_here == -1ULL)
233 panic("Cannot find place for new RAMDISK of size %lld\n",
236 /* Note: this includes all the lowmem currently occupied by
237 the initrd, we rely on that fact to keep the data intact. */
238 reserve_early(ramdisk_here, ramdisk_here + ramdisk_size,
240 initrd_start = ramdisk_here + PAGE_OFFSET;
241 initrd_end = initrd_start + ramdisk_size;
242 printk(KERN_INFO "Allocated new RAMDISK: %08llx - %08llx\n",
243 ramdisk_here, ramdisk_here + ramdisk_size);
245 do_relocate_initrd = true;
248 #define MAX_MAP_CHUNK (NR_FIX_BTMAPS << PAGE_SHIFT)
250 static void __init post_reserve_initrd(void)
252 u64 ramdisk_image = boot_params.hdr.ramdisk_image;
253 u64 ramdisk_size = boot_params.hdr.ramdisk_size;
254 u64 end_of_lowmem = max_low_pfn << PAGE_SHIFT;
256 unsigned long slop, clen, mapaddr;
259 if (!do_relocate_initrd)
262 ramdisk_here = initrd_start - PAGE_OFFSET;
264 q = (char *)initrd_start;
266 /* Copy any lowmem portion of the initrd */
267 if (ramdisk_image < end_of_lowmem) {
268 clen = end_of_lowmem - ramdisk_image;
269 p = (char *)__va(ramdisk_image);
272 /* need to free these low pages...*/
273 printk(KERN_INFO "Freeing old partial RAMDISK %08llx-%08llx\n",
274 ramdisk_image, ramdisk_image + clen - 1);
275 free_bootmem(ramdisk_image, clen);
276 ramdisk_image += clen;
277 ramdisk_size -= clen;
280 /* Copy the highmem portion of the initrd */
281 while (ramdisk_size) {
282 slop = ramdisk_image & ~PAGE_MASK;
284 if (clen > MAX_MAP_CHUNK-slop)
285 clen = MAX_MAP_CHUNK-slop;
286 mapaddr = ramdisk_image & PAGE_MASK;
287 p = early_ioremap(mapaddr, clen+slop);
288 memcpy(q, p+slop, clen);
289 early_iounmap(p, clen+slop);
291 ramdisk_image += clen;
292 ramdisk_size -= clen;
294 /* high pages is not converted by early_res_to_bootmem */
295 ramdisk_image = boot_params.hdr.ramdisk_image;
296 ramdisk_size = boot_params.hdr.ramdisk_size;
297 printk(KERN_INFO "Copied RAMDISK from %016llx - %016llx to %08llx - %08llx\n",
298 ramdisk_image, ramdisk_image + ramdisk_size - 1,
299 ramdisk_here, ramdisk_here + ramdisk_size - 1);
301 /* need to free that, otherwise init highmem will reserve it again */
302 free_early(ramdisk_image, ramdisk_image+ramdisk_size);
305 void __init reserve_initrd(void)
308 static void __init post_reserve_initrd(void)
311 #endif /* CONFIG_BLK_DEV_INITRD */
314 * The node 0 pgdat is initialized before all of these because
315 * it's needed for bootmem. node>0 pgdats have their virtual
316 * space allocated before the pagetables are in place to access
317 * them, so they can't be cleared then.
319 * This should all compile down to nothing when NUMA is off.
321 static void __init remapped_pgdat_init(void)
325 for_each_online_node(nid) {
327 memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
332 static void set_mca_bus(int x)
337 static void set_mca_bus(int x) { }
341 * Determine if we were loaded by an EFI loader. If so, then we have also been
342 * passed the efi memmap, systab, etc., so we should use these data structures
343 * for initialization. Note, the efi init code path is determined by the
344 * global efi_enabled. This allows the same kernel image to be used on existing
345 * systems (with a traditional BIOS) as well as on EFI systems.
347 void __init setup_arch(char **cmdline_p)
349 unsigned long max_low_pfn;
351 memcpy(&boot_cpu_data, &new_cpu_data, sizeof(new_cpu_data));
352 pre_setup_arch_hook();
354 early_ioremap_init();
355 reserve_setup_data();
358 if (!strncmp((char *)&boot_params.efi_info.efi_loader_signature,
365 ROOT_DEV = old_decode_dev(boot_params.hdr.root_dev);
366 screen_info = boot_params.screen_info;
367 edid_info = boot_params.edid_info;
368 apm_info.bios = boot_params.apm_bios_info;
369 ist_info = boot_params.ist_info;
370 saved_video_mode = boot_params.hdr.vid_mode;
371 if( boot_params.sys_desc_table.length != 0 ) {
372 set_mca_bus(boot_params.sys_desc_table.table[3] & 0x2);
373 machine_id = boot_params.sys_desc_table.table[0];
374 machine_submodel_id = boot_params.sys_desc_table.table[1];
375 BIOS_revision = boot_params.sys_desc_table.table[2];
377 bootloader_type = boot_params.hdr.type_of_loader;
379 #ifdef CONFIG_BLK_DEV_RAM
380 rd_image_start = boot_params.hdr.ram_size & RAMDISK_IMAGE_START_MASK;
381 rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
382 rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
390 if (!boot_params.hdr.root_flags)
391 root_mountflags &= ~MS_RDONLY;
392 init_mm.start_code = (unsigned long) _text;
393 init_mm.end_code = (unsigned long) _etext;
394 init_mm.end_data = (unsigned long) _edata;
395 init_mm.brk = init_pg_tables_end + PAGE_OFFSET;
397 code_resource.start = virt_to_phys(_text);
398 code_resource.end = virt_to_phys(_etext)-1;
399 data_resource.start = virt_to_phys(_etext);
400 data_resource.end = virt_to_phys(_edata)-1;
401 bss_resource.start = virt_to_phys(&__bss_start);
402 bss_resource.end = virt_to_phys(&__bss_stop)-1;
408 if (acpi_mps_check()){
409 enable_local_apic = -1;
410 clear_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC);
413 finish_e820_parsing();
417 /* after parse_early_param, so could debug it */
418 insert_resource(&iomem_resource, &code_resource);
419 insert_resource(&iomem_resource, &data_resource);
420 insert_resource(&iomem_resource, &bss_resource);
422 strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
423 *cmdline_p = command_line;
428 if (ppro_with_ram_bug()) {
429 e820_update_range(0x70000000ULL, 0x40000ULL, E820_RAM,
431 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
432 printk(KERN_INFO "fixed physical RAM map:\n");
433 e820_print_map("bad_ppro");
436 e820_register_active_regions(0, 0, -1UL);
438 * partially used pages are not usable - thus
439 * we are rounding upwards:
441 max_pfn = e820_end_of_ram();
443 /* preallocate 4k for mptable mpc */
444 early_reserve_e820_mpc_new();
445 /* update e820 for memory not covered by WB MTRRs */
447 if (mtrr_trim_uncached_memory(max_pfn)) {
448 remove_all_active_ranges();
449 e820_register_active_regions(0, 0, -1UL);
450 max_pfn = e820_end_of_ram();
458 * Parse the ACPI tables for possible boot-time SMP configuration.
460 acpi_boot_table_init();
462 #ifdef CONFIG_ACPI_NUMA
464 * Parse SRAT to discover nodes.
469 max_low_pfn = initmem_init(0, max_pfn);
471 #ifdef CONFIG_ACPI_SLEEP
473 * Reserve low memory region for sleep support.
475 acpi_reserve_bootmem();
477 #ifdef CONFIG_X86_FIND_SMP_CONFIG
479 * Find and reserve possible boot-time SMP configuration:
483 reserve_crashkernel();
485 reserve_ibft_region();
487 #ifdef CONFIG_KVM_CLOCK
493 * Must be after max_low_pfn is determined, and before kernel
494 * pagetables are setup.
501 * NOTE: before this point _nobody_ is allowed to allocate
502 * any memory using the bootmem allocator. Although the
503 * allocator is now initialised only the first 8Mb of the kernel
504 * virtual address space has been mapped. All allocations before
505 * paging_init() has completed must use the alloc_bootmem_low_pages()
506 * variant (which allocates DMA'able memory) and care must be taken
507 * not to exceed the 8Mb limit.
513 * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
516 #ifdef CONFIG_PROVIDE_OHCI1394_DMA_INIT
517 if (init_ohci1394_dma_early)
518 init_ohci1394_dma_on_all_controllers();
522 * NOTE: at this point the bootmem allocator is fully available.
525 post_reserve_initrd();
527 remapped_pgdat_init();
531 paravirt_post_allocator_init();
533 #ifdef CONFIG_X86_GENERICARCH
534 generic_apic_probe();
541 #if defined(CONFIG_X86_MPPARSE) || defined(CONFIG_X86_VISWS)
542 if (smp_found_config)
545 #if defined(CONFIG_SMP) && defined(CONFIG_X86_PC)
547 printk(KERN_WARNING "More than 8 CPUs detected and "
548 "CONFIG_X86_PC cannot handle it.\nUse "
549 "CONFIG_X86_GENERICARCH or CONFIG_X86_BIGSMP.\n");
552 e820_reserve_resources();
553 e820_mark_nosave_regions(max_low_pfn);
555 request_resource(&iomem_resource, &video_ram_resource);
556 reserve_standard_io_resources();
561 #if defined(CONFIG_VGA_CONSOLE)
562 if (!efi_enabled || (efi_mem_type(0xa0000) != EFI_CONVENTIONAL_MEMORY))
563 conswitchp = &vga_con;
564 #elif defined(CONFIG_DUMMY_CONSOLE)
565 conswitchp = &dummy_con;